[
Lists Home |
Date Index |
Thread Index
]
> -----Original Message-----
> From: Tim.Bray@Sun.COM [mailto:Tim.Bray@Sun.COM] On Behalf Of Tim Bray
> Sent: Monday, April 12, 2004 18:25
> To: xml-dev@lists.xml.org; DEV
> Subject: Re: [xml-dev] XML Binary Characterization WG public
> list available
>
>
> On Apr 12, 2004, at 1:58 PM, Santiago Pericas-Geertsen wrote:
>
> > A completely different spec is X.finf (finf = Fast
> Infoset) in which
> > a schema for the XML infoset is being defined. This spec will allow
> > encoding arbitrary XML documents without the need of a
> schema. In the
> > future it may also be possible to use X.finf together with X.694 to
> > encode open content (e.g. xsd:any).
>
> Right, but I thought the whole X.694 idea was that because
> you knew the
> schema you could encode character data that was representing numbers
> super-efficiently and represent the tags as small integers; thus the
> saving. So does X.finf actually give you that much of a saving?
There were some preliminary test results included in OSS Nokalva's position
paper at last year's W3C binary workshop. Those result were based on an
initial version of the ASN.1 schema for the infoset (now being called "fast
infoset").
The measured reduction in size from the original XML was interesting (19% to
69% using Aligned PER). The measured speed was always higher than the speed
of Xerces/C 2.2 parsing the XML (although not dramatically higher). Anyway,
the initial ASN.1 schema for the infoset was certainly faster than Xerces/C
2.2 for all samples, so it apparently provided both size and speed benefits
over the original XML.
The "fast infoset" has been improved since then, and still is, and we expect
the final version to be faster than what the results of that early test
indicated.
Alessandro Triglia
OSS Nokalva
> -Tim
>
|
